1. Field of the Invention
The present invention relates generally to computer network systems and, more particularly, to management of computer network virtualization environments.
2. Description of the Related Art
Information Technology (IT) management tasks can be characterized into two general areas, managing present day operations and forecasting capacity for future operations. Ensuring the well being of the current IT environment while using current trends to predict needs and trends for the future business needs is a fine balance and a highly refined skill Management needs all the help and tools it can find to assist it in these tasks. Today, all major IT management platforms support the International Telecommunications Union (ITU) standard for Element Management Systems (EMS), wherein general functionality can be split into five key areas: Fault, Configuration, Accounting, Performance, and Security (FCAPS).
This conventional methodology has created an element-driven management system, with a focus on ensuring that each of the individual elements are running to their full potential. As the number of elements grew, the need for aggregated and correlated information increased. As the number of data-center locations grew, the need for global visibility and control increased. Conventionally, IT capacity planning for day-to-day operations is typically carried out with a bottom-up data aggregation and with the use of forecasting methods such as trending, simulation, and custom analytics.
Capacity planning for resources is also typically completed when new business applications are rolled out or during an application upgrade cycle. In this capacity planning scenario, the planning is typically carried out at the individual device level, which is then multiplied by the number of consumers and/or producers and further multiplied by the number of locations that need to be supported, giving a large number in the aggregate: (number of individual devices)×(number of consumers/producers)×(number of locations requiring support).
In order to estimate what resource capacity an enterprise will need to support the core business applications it provides, an enterprise will typically evaluate the worst case usage scenario, and bolster its capacity to ensure that a worst case scenario will be adequately supported. What is often overlooked is that this type of worst-case capacity planning typically is driven by the vendors who have built ROI calculators that are to their benefit.
By planning and bulking up resources to combat this evasive worst case scenario, enterprises typically end up with under-utilized IT resources. Applying the Pareto Principle, an estimate of how much under-utilized capacity can exist in a single enterprise would be as follows: only 20% of the available capacity ends up being used during 80% of the time during a given timeframe.
Many improvements have been made in the enterprise scenario for management of IT resources. At the macro-level, IT resources can be classified into 4 categories: (1) client resources (client resource examples include desktop machines, wireless, handheld devices, client grids (such as SETI@home and the like); (2) server resources (server resource examples include mainframe, File Server, Web server, peer-to-peer servers, blade servers, grid servers etc.); (3) network resources (network resource example include routers, switches, bridges, infiniband, wireless, radio, optical, fiber channel, link aggregation technologies (such as BitTorrent and the like); (4) storage resources (storage resource example include databases, network attached storage, storage area networks, data grids etc.). While one can describe a capacity planning scenario for each of the categories above, they all follow a very similar capacity planning process.
In the following example, we will describe a typical server resource capacity planning scenario. Servers in the enterprise have evolved with new application architecture. Application topologies have evolved from Mainframe-Green Screen interaction, to Client/Server, to Client/Web Server/Application Server/Database, Peer-to-Peer, and so forth. Server resource capacity planning is typically achieved by stress-testing the application with a certain predetermined workload and a set, acceptable application response time. A hardware specification is defined to support a user-defined worst-case scenario. The application is rolled out on the new hardware into a production environment.
Server resource and application utilization is monitored by a FCAPS-compliant management platform to provide complete visibility over operations. In order to provide an aggregated summary view, such management platforms typically roll-up element-level metrics into higher level metrics through data correlation techniques.
Conventionally, with the emergence of resource virtualization and the increase in use of Web services, combined with service-oriented architectures (SOAs), the number of moving pieces that need to be managed for the enterprise continues to rise. For example, imagine an enterprise running composite applications. That enterprise would include using a mixture of legacy, local, and external Web services, running on virtual infrastructure spread globally across the enterprise, and frustrated end users can't complete their mission critical business tasks. It is difficult to achieve sufficient visibility and control to manage such an environment, and knowing where to begin to manage such an environment can be difficult.
Traditional methods of resource planning at the individual physical resource level begin to show their age. For example, correlation and aggregation of element level data also becomes compute-intensive with the increase in the number of managed elements. It has been said that “The information technology industry is in a strange situation. We have enormously sophisticated engines that we're running—in the form of CPUs and communications equipment and so forth, but the way we keep them running is through an outdated vision.” Doug Busch, Intel Vice President and CIO-Technology quoted in Intel Magazine, September/October 2004 (available at the URL of: www.intel.com/update/contents/it09041.htm as of June 2005). Management of virtual assets can be achieved conventionally with virtualization software tools, but such techniques are typically labor intensive and require manual selection and implementation of configurations and utilize relatively cumbersome configuration change management.
In network systems, with virtualization, it is possible to deploy physical resources in the form of virtual assets. The assets can thereby provide the functional equivalent of desktops (user interfaces), operating systems, applications, servers, data bases, and the like. Adding additional applications can be implemented by remotely executed software operations in virtual environments on one or more computers, rather than physical installations involving personnel with an installation CD media at each physical location (computer) of a network where the additional applications are desired.
The management of such virtual assets, however, is becoming increasingly complex and unwieldy. Many tools to assist in the management of virtualization environments are proprietary and work only with virtual environments from particular vendors. Similarly, some virtualization tools might only work with specific central processor units (CPUs) of machines that host the virtual environment, or might only work with specific operating systems or virtualization platforms of either the host machine or in the virtual environment. This characteristic can make it necessary to have multiple tools on hand for the various platforms and vendors that might be deployed throughout a network, as well as making it necessary to acquire and maintain the skill sets necessary to use such tools. The mere fact of requiring such diverse tools is, itself, inefficient. Thus, although virtualization trends show much promise for more efficient utilization of physical resources by optimal deployment of virtual assets, the virtualization environment management task is daunting.
It would be advantageous if more efficient means for managing virtual environments across computer networks were available. The present invention satisfies this need.